Anchored logistic agent-based model of cooperation among fourteen European Union member states, with structural diagnostics 1.1.0
An agent-based model in which fourteen European Union member states each carry a cooperation
propensity in [0,1], updated by a logistic link applied to a latent index that combines the unit’s
own previous state, a weighted average of the other units’ states, and four min-max normalised
exogenous indicators (cooperative benefit, opportunistic temptation, systemic exit cost,
intertemporal confidence). The logistic centres are anchored so that the observed 2024
configuration is an exact fixed point of the baseline map, the device that makes such models
“empirically calibrated”. Scenarios apply multipliers to the indicators over a 2024-2040 horizon.
The model is distributed here as a diagnostic object rather than as a predictive instrument. It
accompanies a methods paper that proves three structural results about this class of models and
uses this implementation as the worked example: the one-step map is a contraction on the unit cube
whenever k(rho + eta) < 4, for every row-stochastic interaction matrix and every scenario, so
gradual and ordered degradation is a property of the map rather than a finding about the system;
the empirical anchor loses stability at a computable slope, with a fold just below it beyond which
shock paths become irreversible; and under min-max normalisation with multiplicative scenarios the
perturbation a unit receives is proportional to its distance from the sample minimum, so
distributional results are confounded with sample membership.
The package includes the model, the three diagnostics, the scripts that produce both figures, a
repaired specification at four cumulative levels, and a script that re-checks every numerical claim
in the accompanying paper against a computed value (59 checks). The worked example is documented
following the ODD protocol.
Release Notes
First public release. Accompanies a methods paper on the structural limits of anchored logistic agent-based models, where this model serves as the fully specified worked example. Includes the model, three diagnostics (contraction modulus and critical slope; stability threshold of the empirical anchor with a multi-start fixed-point search and a hysteresis sweep; shock incidence under min-max normalisation with a leave-one-out recomputation), a repaired specification at four cumulative levels, the scripts producing both figures, and a verification script that re-checks every numerical claim in the paper against a computed value (67 checks with the data files present, 59 without). Tested on Python 3.11